Resource Allocation Scheme Based on Deep Reinforcement Learning for Device-to-Device Communications

Abstract

In this paper, we propose a decentralized resource allocation scheme based on deep reinforcement learning designed for device-to-device communications underlay cellular networks. The proposed scheme allocates appropriate channel resource and transmit power to each D2D pairs iteratively to maximize the overall effective throughput by utilizing observation consisting of location information of mobile devices and resource allocation of the other devices.